Acoustics of rubbing feathers: the velvet of owl feathers reduces frictional noise

Abstract

One feather structure associated with an owl's ability to fly quietly is the soft dorsal surface on their flight feathers: the velvet. This velvet is a mat of elongated filamentous pennulums that extend up from feather barbules. The aerodynamic noise hypothesis posits this velvet reduces aerodynamic noise caused by the formation of turbulence, while the structural noise hypothesis posits the velvet acts as a dry lubricant, reducing frictional noise produced by feathers sliding past one another. We investigated the structural noise hypothesis by quantifying the length of the velvet on 24 locations across the wing of the barred owl (Strix varia) and then qualitatively assessing the presence of velvet in 24 bird species. We found that velvet has evolved at least 4 times independently (convergently) in owls, nightbirds, hawks and falcons. Then, we rubbed 96 pairs of feathers together from 17 bird species (including the four clades that have independently evolved velvet) under three experimental treatments: control, hairspray applied (to impair the velvet) and hairspray removed. The sound of feathers rubbing against each other was broadband, similar to the sound of rubbing sandpaper or Velcro. Species with velvet produced rubbing sounds that were 20.9 dB quieter than species without velvet, and velvet-coated feathers became 7.4 dB louder when manipulated with hairspray, while feathers lacking velvet only increased in loudness by 1.7 dB, relative to the control treatments. These results all support the hypothesis that the velvet primarily functions to ameliorate the sounds of feathers rubbing against other feathers.

Keywords: Aeroacoustics; Locomotion-induced sound; Pennulums; Quiet flight; Sonation; Strigiformes.

Fig. 1.

Owl feather and velvet anatomy. (A) Velvet on the surface of barred owl (Strix varia) feather p4 (location: inner vane, halfway down the length of the vane), ×30 magnification. (B) Vane of a feather that lacks the velvet (Hawaiian goose, Branta sandvicensis, p4, ×30 magnification). (C) Anatomical features of a single barb from a barred owl feather, showing a distal barbule with elongated pennulum, ×30 magnification. (D) Cross-section of a barb, showing the distal barbule, proximal barbule, hooklets (which attach the distal barbule to the neighboring proximal barbule), and elongated pennulum that makes up the dorsal velvet of owl feathers. Drawn by L.G.L. following Lucas and Stettenheim (1972). (E) Individual flight feathers are composed of an intricate branching system stemming from the rachis – the central shaft that barbs are attached to; they are further joined by smaller barbules projecting off it. Each barb has rows of barbules attached: the proximal barbules (bow radiates) all point towards the base of the feather while the distal barbules (hook radiates), often composed of multiple hooklets, point towards the tip of the feather. Each barbule has two segments: a base portion and a pennulum portion. The rachis divides the whole feather into the inner vane and outer vane.

Fig. 2.

Feather rubbing experimental design and preliminary results. (A) The overlapping feather was affixed to a turntable and rotated with the turntable so that it slid against the underlapping feather, with a microphone 9 cm away from the midpoint of the overlapping feather region. (B) Speed of rotation of the turntable had a relatively small effect on loudness of rubbed barred owl (solid lines) and Hawaiian goose (dashed lines) feathers, with a least squares linear regression model effect size of 5.75 rad s⁻¹ dB⁻¹ (n=2 species, n=2 feather pairs, 6 measurements per pair per treatment). (C) In the second experimental set up, feathers were held by hand, with the overlapping region midpoint positioned 10 cm away from the microphone. While the underlapping feather remained motionless, the overlapping feather was pulled away from the underlapping feather (starting from the overlapped position) in the horizontal plane.

Fig. 3.

Effects of adding hairspray on rubbing sounds of 17 species of bird. (A) Amplitude (SPL, sound pressure level) of rubbing sounds across three experimental treatments: no hairspray (treatment 1; blue), hairspray applied (treatment 2; red) and hairspray removed (treatment 3; green). The binary presence (black squares) or absence (white squares) of velvet on flight feathers is indicated below. Box plots show 25th, median and 75th percentiles; whiskers are 1.5 times the interquartile range. See Table 2 for statistics. Phylogeny pruned from Prum et al. (2015). (B) Spectra from barred owl P5 rubbed against P6. Measured from 0.15 s of sound. Note: y-axis not adjusted for microphone gain settings.

Fig. 4.

Power spectra of sound recordings of feathers rubbing together, whooshing, Velcro being pulled apart and wing flapping of a chickadee. (A) Power spectra (kHz, log scale) of rubbing sounds from seven species of bird (out of 17 measured in this experiment). Five species naturally have velvet (dark blue) while two species lack velvet (light blue). N=7 species measured, N=1 feather pairs per species. Note: y-axis not adjusted for microphone gain settings. (B) Spectrograms of rubbing trials. Blue and red boxes correspond to power spectra of rubbing (blue) and background sound (red) as shown in A. See Fig. 2C for experimental setup. Note that the microphone (Brüel & Kjær 4190) had flat sensitivity between 1 and 20 kHz, and rolled off above 20 kHz. (C) Aerodynamic whooshing sound produced by a condor primary feather swung rapidly past the microphone, where most aerodynamic sound is below 6 kHz, N=1 feather. (D) Sound of two strips of Velcro being pulled apart, so the hooks and loops detach from one another. Velcro produces broadband sound with approximately equal sound energy at all frequencies (white noise) between 1 and 20 kHz. N=1. (E) Black-capped chickadee flapping sounds (supplemental data from Fournier et al., 2013), recorded with an ultrasound (200 kHz) microphone. Wing flapping sounds include substantial energy up to approximately 70 kHz, N=1. For C–E, frequency is shown on a log scale.

Fig. 5.

Velvet measurements on feathers of the barred owl (Strix varia). Four feathers, p10, p8, s3 and s8, from N=3 barred owls (both wings, i.e. N=6 feathers), had 6 individual distal barbules measured at three positions per feather. Blue: barbules exposed to the boundary layer on the wing, green: barbules expected to be covered by and rub against an overlapping feather (means±s.d.). Raw data are in Table S3.